Senior Associate: Agentic AI & DevOps Engineer

PwC

Monterrey

Presencial

MXN 600.000 - 900.000

Jornada completa

14 días+

Recibe más respuestas de empleadores

Envía un currículum específico para el puesto de trabajo en cuestión de minutos.

Descripción de la vacante

PwC Acceleration Center Mexico seeks an Acceleration Center Agentic AI Engineer — Senior Associate to transform raw data into actionable insights and drive business growth across client portfolios, leveraging Agentic AI, RAG, and LLMs. You will oversee model layer end-to-end, from prompt design to MLOps, ensuring scalable, cost-efficient production.

You will work with global teams across Advisory, Assurance, Tax and Business Services, building observability, evaluation frameworks, and CI/CD for

Formación

At least a Bachelor’s degree.
At least 2 years of experience.
Oral and written proficiency in English required.

Responsabilidades

Design and deploy LLM-powered automations at scale - agents, skills, pipelines - from prototype through production.
Build eval frameworks: offline benchmarks, adversarial red‑teaming, regression suites, and production feedback loops.
Own observability end-to-end: distributed traces, latency histograms, cost attribution, token throughput, error budgets, and dashboards that surface signal over noise.
Define and enforce SLOs for model-backed services; own the alerting stack and incident response playbooks.
Debug production failures across the full stack - from Kubernetes pod restarts to model hallucinations to prompt injection - and write postmortems that actually prevent recurrence.
Operate AI-natively: use Claude Code, Codex, GitHub Copilot, Cursor, and similar tools as daily infrastructure, not novelty - and hold the output to the same standard as hand‑written code.
Instrument and monitor LLM behavior in production: token usage, response quality, failure modes, latency SLAs Architect CI/CD for model and prompt changes: versioning, automated eval gates, rollback strategies.
MLOps/SRE: shipped LLM‑backed automations to productions with real SLAs and real users, reproducible experiments (tracked, versioned, auditable), SLO definition, error tracking, alerting design, monitoring pipelines for model degradation, prompt drift and cost anomalies.
AI Native Fluency: daily user of AI coding assistants (Claude Code, Codex, Copilot, Cursor), understand in‑model limitations, failure modes and when to distrust output, can write prompts and agent instructions with rigor, has opinions on eval design, context management and tool‑use patterns.